Unknown critical error on Fedora 31

[Amarok19]

I encounter an error when trying to run a Windows 10 virtual machine on a Fedora 31 host. The machine was working fine earlier, I suppose the issue came with some update (host kernel or the VirtualBox package itself) but I have to little expertise to draw any conclusions from log analysis. (I'm attaching the log file under a link included in an attached file -- I've only just established an account and I have no other way to circumvent the forum restrictions and at the same time this post would make zero sense without a log file included...) The machine simply enters Guru meditation right after I start it.

[Martin]

Please zip the logfile and attach it here directly.

[scottgus1]

attach it here directly.

This is done with the forum's Upload Attachment tab. Available from the first post by any new poster, not blocked by the no-url filter.

[Amarok19]

I hope 7z file is fine (an ordinary .zip still exceeds the 256 KiB limit)

[Martin]

A single logfile would have been sufficient.
Could you try disabling 3D support for a test?
The VM throws a Guru Meditation directly after initialazind VMSVGA3d with some VERR_INVALID_PARAMETER errors.

Code: Select all

00:00:00.727443 VMSVGA3d: cap[77]=0x00000100 {MAX_CONTEXT_IDS}
00:00:00.727444 VMSVGA3d: cap[78]=0x00008000 {MAX_SURFACE_IDS}
00:00:00.727444 VMSVGA3d: cap[79]=0x000000c1 {SURFACEFMT_Z_DF16}
00:00:00.727445 VMSVGA3d: cap[80]=0x000000c1 {SURFACEFMT_Z_DF24}
00:00:00.727446 VMSVGA3d: cap[81]=0x000000c1 {SURFACEFMT_Z_D24S8_INT}
00:00:00.727447 VMSVGA3d: cap[82]=failed rc=VERR_INVALID_PARAMETER! {SURFACEFMT_BC4_UNORM}
00:00:00.727448 VMSVGA3d: cap[83]=failed rc=VERR_INVALID_PARAMETER! {SURFACEFMT_BC5_UNORM}
00:00:00.727457 AIOMgr: Endpoints without assigned bandwidth groups:
00:00:00.727492 AIOMgr:     /home/setlakk/VirtualBox VMs/Windows 10/Windows 10.vdi
00:00:00.727765 Changing the VM state from 'POWERING_ON' to 'RUNNING'
00:00:00.727776 Console: Machine state changed to 'Running'
00:00:00.729321 Changing the VM state from 'RUNNING' to 'GURU_MEDITATION'
00:00:00.729343 Console: Machine state changed to 'GuruMeditation'
00:00:00.730320 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
00:00:00.730321 !!
00:00:00.730321 !!         VCPU0: Guru Meditation -79 (VERR_INVALID_STATE)

Also
00:00:00.570896 NumCPUs <integer> = 0x0000000000000004 (4)
00:00:00.697746 CPUM: Physical host cores: 4
could you try to set the VM to 2 vCPU?
The host needs "free" CPU cores to handle the i/o of running VMs.

[Amarok19]

I've changed these two parameters... Still no luck.

[Amarok19]

... and surprisingly, today, fresh after booting the PC, it works. I have no idea what changed, I did not touch VM settings and no significant updates arrived. It just started working.

[fth0]

The problem occurs very early, when the 2 vCPUs haven't executed any CPU instruction yet (BIOS reset vector). If it occurs again, I suggest to try two things (separately and/or in combination):

• If you have qemu-kvm installed on your Linux host, remove it.
• In VM settings > System > Processor, disable Nested VT-x/AMD-V.

[Amarok19]

Thanks for the suggestion! If the problem occurs again, I'll try these solutions and post the results.

[Amarok19]

Both actions listed above (in combination) didn't work.

[Amarok19]

Some qemu process was still running on my machine. I've killed it and rebooted the host and now it works.

I suppose it was the only reason - it may cause the problem all the time, so even without deleting qemu-kvm it could work. I would just have to watch running processes and not attempt runnnig a VirtualBox VM when a (probably zombie-like) qemu-based process is running.